[Bug c++/45267] [4.5 regression] inlining fails with -m32

["rguenth at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

------- Comment #9 from rguenth at gcc dot gnu dot org  2010-08-12 22:16 -------
116             {
1117              if (!cgraph_recursive_inlining_p (edge->caller, edge->callee,
1118                                                &edge->inline_failed))
1119                {
1120                  edge->inline_failed = not_good;
1121                  if (dump_file)
1122                    fprintf (dump_file, " inline_failed:%s.\n",
1123                             cgraph_inline_failed_string
(edge->inline_failed));
1124                }
1125              continue;
(gdb) p not_good
$13 = CIF_UNLIKELY_CALL
(gdb) p edge->callee->local.disregard_inline_limits 
$16 = 0
(gdb) p edge->callee->decl->function_decl.disregard_inline_limits 
$17 = 0
(gdb) call debug_tree ( edge->callee->decl->decl_common.attributes)
 <tree_list 0xb779da68
    purpose <identifier_node 0xb77a4784 always_inline
    bindings <(nil)>
    local bindings <(nil)>>>

the flag is cleared from the template instantiation functin decl here:

  if (TREE_CODE (newdecl) == FUNCTION_DECL)
    {
      int function_size;

      function_size = sizeof (struct tree_decl_common);

      memcpy ((char *) olddecl + sizeof (struct tree_common),
              (char *) newdecl + sizeof (struct tree_common),
              function_size - sizeof (struct tree_common));

      memcpy ((char *) olddecl + sizeof (struct tree_decl_common),
              (char *) newdecl + sizeof (struct tree_decl_common),
              sizeof (struct tree_function_decl) - sizeof (struct
tree_decl_common));

overwritten from a variant built via

#0  0x0811eb3e in grokfndecl (ctype=0xb77a64e0, type=0xb77a67e0, 
    declarator=0xb76f0888, parms=0xb77aa230, orig_declarator=0xb76f0888, 
    virtualp=0, flags=NO_SPECIAL, quals=1, raises=0x0, check=1, friendp=0, 
    publicp=1, inlinep=1, sfk=sfk_none, funcdef_flag=1 '\001', 
    template_count=1, in_namespace=0x0, attrlist=0xbffff088, location=2025)
    at /home/richard/src/gcc-4_5-branch/gcc/cp/decl.c:6655
#1  0x0812c65c in grokdeclarator (declarator=0x90a7814, declspecs=0xbffff19c, 
    decl_context=NORMAL, initialized=1, attrlist=0xbffff088)
    at /home/richard/src/gcc-4_5-branch/gcc/cp/decl.c:9586
#2  0x081466c4 in start_function (declspecs=0xbffff19c, declarator=0x90a78b8, 
    attrs=0x0) at /home/richard/src/gcc-4_5-branch/gcc/cp/decl.c:12096
#3  0x0826a771 in cp_parser_function_definition_from_specifiers_and_declarator
    (parser=0xb77a4854, decl_specifiers=0xbffff19c, attributes=0x0, 
    declarator=0x90a78b8)
    at /home/richard/src/gcc-4_5-branch/gcc/cp/parser.c:18548

we can again check the attribute lists all the time, but really
DECL_DISREGARD_INLINE_LIMITS should be kept valid here.

We already merge that in various places in duplicate_decls, but appearantly
not here.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at redhat dot com
          Component|middle-end                  |c++


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45267